1. A multi-completion method of motor vehicle exhaust telemetering data based on a space-time convolution network is characterized by comprising the following steps:

step S1: acquiring topological structure information of remote measurement points of a road network in a detection area and multiple missing data of the exhaust concentration of the motor vehicle remotely measured by remote measuring equipment in the road network, processing the topological structure information and the multiple missing data, and respectively constructing topological structure chart data of a spatial road network and time series data of the exhaust concentration of the motor vehicle;

step S2: inputting the topological structure diagram data of the space road network and the time sequence data of the exhaust concentration of the motor vehicle into a space-time diagram convolution network for completion, wherein the space-time diagram convolution network comprises: a self-attention mechanism space-time graph convolution network and an improved nearest neighbor algorithm; and inputting the topological structure diagram data of the space road network into the space-time diagram convolutional network of the self-attention mechanism to obtain a space-time characteristic result of the road network, and using the space-time characteristic result in the improved nearest neighbor algorithm to complete multiple missing data of the motor vehicle exhaust emission concentration time sequence missing data to obtain final complete motor vehicle exhaust emission concentration time sequence data.

2. The space-time convolutional network-based multi-completion method for vehicle exhaust telemetry data as claimed in claim 1, wherein the step S1: acquiring topological structure information of remote measuring points of a road network in a detection area and multiple missing data of the exhaust concentration of the motor vehicle remotely measured by remote measuring equipment in the road network, processing the topological structure data and the time series data of the exhaust concentration of the motor vehicle in the space road network, and respectively constructing the topological structure data and the time series data of the exhaust concentration of the motor vehicle in the space road network, wherein the topological structure data comprises the following steps:

step S11: acquiring topological structure information of remote measuring points of a spatial road network of a detection area, and constructing topological structure chart data of the spatial road network;

step S12: the method comprises the steps of collecting motor vehicle exhaust emission concentration multiple missing data telemetered by telemetering equipment in a road network, and constructing exhaust concentration emission time sequence input data of multiple time periods.

3. The space-time convolutional network-based multi-completion method for vehicle exhaust telemetry data as claimed in claim 2, wherein the step S11: acquiring topological structure information of remote measuring points of a spatial road network in a detection area, and constructing topological structure diagram data of the spatial road network, wherein the topological structure diagram data comprises the following steps:

step S111: acquiring longitude and latitude in a map and approximate distances between detection points in road network remote sensing detection nodes in a detection area, wherein the number of the remote sensing detection nodes is N, and the number of the nodes v is_iAnd v_jApproximate geographical distance between is l_i,jWherein i, j is epsilon to N;

step S112: constructing topological structure diagram data of a spatial road network, and defining a topological structure diagram G of the spatial road network as (V, E, A) according to a road network remote sensing detection node topological diagram, wherein G is an undirected graph; v is a node set for remote sensing detection of tail gas emission of the motor vehicle, | V | ═ N is the number of nodes, E is an edge set of remote sensing points, the connectivity of the nodes is represented, A ∈ R^N×NThe weight adjacency matrix of the graph G has a matrix A element mathematical expression shown in a formula (1):

wherein link (i, j) represents the node v_iAnd v_jThe road connection between them; when link (i, j) is 1, it represents the node v_iAnd v_jThere is a road connection between them; the parameter θ represents the size and sparsity of the control weight adjacency matrix.

4. The space-time convolutional network-based multi-completion method for vehicle exhaust telemetry data as claimed in claim 1, wherein the step S2: inputting the topological structure diagram data of the space road network and the time-series data of the exhaust concentration of the motor vehicle into a space-time diagram convolution network for completion, wherein the space-time diagram convolution network comprises a self-attention mechanism space-time diagram convolution network and an improved nearest neighbor algorithm; inputting the topological structure diagram data of the space road network into the space-time diagram convolutional network of the self-attention mechanism to obtain a space-time characteristic result of the road network, and using the space-time characteristic result in the improved nearest neighbor algorithm to perform multiple missing data completion on the missing data of the time sequence of the exhaust concentration of the motor vehicle exhaust to obtain final complete time sequence data of the exhaust concentration of the motor vehicle exhaust, wherein the method comprises the following steps:

step S21: inputting the topological structure diagram data of the space road network into the self-attention mechanism space-time diagram convolution network to obtain the space-time characteristic weight of the self-attention space-time diagram convolution of the space road network;

step S22: and inputting the convolved space-time characteristic weight of the self-attention space-time diagram and the missing data of the time sequence of the motor vehicle exhaust emission concentration into an improved nearest neighbor algorithm to obtain final complete time sequence data of the motor vehicle exhaust emission concentration.

5. The space-time convolutional network-based multi-completion method for vehicle exhaust telemetry data as claimed in claim 4, wherein the step S21: inputting the topological structure diagram data of the spatial road network into the self-attention mechanism space-time diagram convolution network to obtain the space-time characteristic weight of the self-attention space-time diagram convolution of the spatial road network, wherein the space-time characteristic weight comprises the following steps:

step S211: transforming the space road network topological structure diagram G (V, E, A) from the diagram to the spectrum domain to realize diagram convolution, and using a convolution kernel G_θPerforming convolution operation on the graph G to obtain a space graph convolution, and performing approximate expansion by using a Chebyshev polynomial to obtain a space graph convolution result H, wherein the mathematical expression of the space graph convolution result H is shown as a formula (2):

wherein_GWhich represents a convolution operation, the operation of the convolution,k order Chebyshev polynomial coefficients for K (K ∈ K); x represents a weight adjacency matrix a of the input graph G;λ_maxmaximum eigenvalue, I, of Laplace matrix representing matrix A_NIs the identity matrix of matrix A, and the Laplaca of graph GNormalized form of the sigmoid matrixRecursive definition of chebyshev polynomials: t is_k(x)＝2xT_k-1(x)-T_k-2(x) Wherein, T₀(x)＝1，T₁(x)＝x；

Step S212: calculating a weight coefficient alpha which is given to the spatial map convolution result by the self-attention mechanism according to the following formula (3) by using the self-attention mechanism method for the spatial map convolution result H, and calculating to obtain a self-attention time characteristic result H_Att＝α·H；

Wherein H_i,jThe j value of the ith telemetering device telemetering of H is represented, H is the hours of collection in the motor vehicle telemetering exhaust emission time sequence data set, d is the number of days of collection, and w is the number of weeks of collection;

step S213: the attention time characteristic result H_AttThrough the calculation of at least one layer of time convolution neural network, the space-time characteristic result obtained by output isElements contained therein

6. The space-time convolutional network-based multi-completion method for automotive exhaust telemetry data as claimed in claim 4, wherein the modified nearest neighbor algorithm in the step S22 specifically comprises:

step S221: for exhaust emission data sets containing missing valuesAnd the result of the spatio-temporal featureGiven improved nearest neighbor parameter k, T,And k inputting the modified nearest neighbor algorithm;

step S222: by missing value x in T_i,jCentering, wherein i belongs to N, j belongs to (h + d + w), constructing a root node corresponding to a hyper-rectangular region containing Nx (h + d + w) space missing the exhaust emission data T; the space-time characteristic result is obtainedSequencing the values in the distance data table from small to large to obtain a sequencing Array, sequentially taking k values from the Array, and taking a space-time characteristic result h of the k values_i,jCorresponding motor vehicle exhaust emission data x_i,jAccording to the following formula (4), summing and averaging are carried out to obtain the missing value of the exhaust emission concentration of the position

Step S223: repeating the step S222 to finally obtain a multiple deficiency value exhaust emission data set with complete deficiency value complementation

7. A multi-completion system of motor vehicle exhaust telemetering data based on a space-time convolution network is characterized by comprising the following modules:

the system comprises a space road network data and exhaust emission data building module, a data acquisition module and a data processing module, wherein the space road network data and exhaust emission data building module is used for acquiring topological structure information of road network remote measurement points in a detection area and multiple missing data of exhaust emission concentration of the motor vehicle remotely measured by remote measuring equipment in a road network, processing the topological structure data and the time series data of the exhaust emission concentration of the motor vehicle of the space road network are respectively built;

the space-time graph convolution network module is used for inputting the topological structure diagram data of the space road network and the time sequence data of the exhaust emission concentration of the motor vehicle into the space-time graph convolution network for completion, wherein the space-time graph convolution network comprises: a self-attention mechanism space-time graph convolution network and an improved nearest neighbor algorithm; and inputting the topological structure diagram data of the space road network into the space-time diagram convolutional network of the self-attention mechanism to obtain a space-time characteristic result of the road network, and using the space-time characteristic result in the improved nearest neighbor algorithm to complete multiple missing data of the motor vehicle exhaust emission concentration time sequence missing data to obtain final complete motor vehicle exhaust emission concentration time sequence data.

Background

With the increase of the quantity of urban motor vehicles, the emission of motor vehicle exhaust is increased rapidly, i.e. the natural and environmental problems caused by the emission of urban exhaust are becoming more serious, and a major social problem is brought, so that the monitoring work of motor vehicle exhaust pollution faces a serious risk. Greenhouse gases generated by the exhaust emission of motor vehicles are the main source of urban air pollution and can cause certain harm to human health. Especially, the motor vehicles with exhaust emission exceeding the standard caused by the old unqualified emission and the motor vehicle problems can generate a large amount of gases harmful to the atmospheric environment, and effective supervision on the exhaust emission exceeding the standard is necessary.

Although the domestic telemetry technology has begun to be popularized, the acquired motor vehicle exhaust emission data is easy to be lost in consideration of the complexity of the telemetry equipment and the interference of the field environment, and the lost data can cause the problems of subsequent analysis of the exhaust emission and the like to cause distortion on the source. Most of the existing data value data completion adopts a mean value, mode or median method, or adopts a regression mode to complete single missing data, and the methods only roughly fill missing value parts or only fill single data, and cannot effectively fill multiple missing data sets. Meanwhile, the physical connection and the time dependence of telemetering data exist among all remote measuring points in the detection area, and huge completion errors can be caused when the factors are not considered to complete the data. Therefore, the existing method for complementing the exhaust emission concentration missing data cannot effectively complement the exhaust emission missing data.

Disclosure of Invention

In order to solve the technical problems, the defects of the prior art are overcome. The invention provides a multi-completion method and a multi-completion system for motor vehicle exhaust telemetering data based on a space-time convolution network. Based on the existence of space dependency and time dependency among all the telemetering equipment points of the detection area, the invention introduces the influence of road network space topological structure information and exhaust emission data in a periodic time period in an exhaust emission data time period on the missing data concentration, so that the invention can obtain better completion effect on the motor vehicle exhaust emission concentration missing data set.

The technical solution of the invention is as follows: a multi-completion method of motor vehicle exhaust telemetering data based on a space-time convolution network comprises the following steps:

step S1: acquiring topological structure information of remote measurement points of a road network in a detection area and multiple missing data of the exhaust concentration of the motor vehicle remotely measured by remote measuring equipment in the road network, processing the topological structure information and the multiple missing data, and respectively constructing topological structure chart data of a spatial road network and time series data of the exhaust concentration of the motor vehicle;

step S2: inputting the topological structure diagram data of the space road network and the time sequence data of the motor vehicle exhaust emission concentration into a space-time diagram convolution network for completion, wherein the space-time diagram convolution network comprises a self-attention mechanism space-time diagram convolution network and an improved nearest neighbor algorithm, the topological structure diagram data of the space road network is input into the self-attention mechanism space-time diagram convolution network to obtain a space-time characteristic result of the road network, and the space-time characteristic result is used for performing multiple missing data completion on the missing data of the motor vehicle exhaust emission concentration time sequence in the improved nearest neighbor algorithm to obtain the final complete time sequence data of the motor vehicle exhaust emission concentration.

Compared with the prior art, the invention has the following advantages:

1. the invention provides a multiple completion method and a multiple completion system for motor vehicle exhaust telemetering data based on a space-time convolution network, which are based on a topological structure of detection area telemetering equipment and exhaust emission time sequence data in a certain time period. Selecting topological structure information of a detection area to extract space-time characteristics in the detection area; and then, a plurality of time interval periods (every hour, every day and every week) on the time axis of the tail gas time sequence data are processed by combining the space-time characteristics to form time sequence missing data of tail gas emission for data completion, so that the accuracy of the completion result is improved.

2. The multiple completion method and the multiple completion system of the motor vehicle tail gas remote measurement data based on the space-time convolution network can complete remote measurement of motor vehicle tail gas emission concentration missing data. Space-time graph convolution and a self-attention mechanism are used for extracting space and time dependence for the first time, space-time characteristics with weights are generated, and exhaust emission concentration data completion is carried out by using an improved nearest neighbor method based on space-time characteristic results.

Drawings

FIG. 1 is a flow chart of a method for multi-complementing motor vehicle exhaust telemetry data based on a spatio-temporal convolutional network in an embodiment of the present invention;

fig. 2 is a block diagram of a multi-completion method for vehicle exhaust telemetry data based on a space-time convolutional network according to an embodiment of the present invention, in which step S1: acquiring topological structure information of remote measurement points of a road network in a detection area and multiple missing data of the exhaust concentration of the motor vehicle remotely measured by remote measuring equipment in the road network, processing the topological structure information and the multiple missing data, and respectively constructing a topological structure chart data of a spatial road network and a flow chart of time series data of the exhaust concentration of the motor vehicle;

FIG. 3 shows a step S2 of the multi-completion method for the vehicle exhaust telemetry data based on the spatio-temporal convolution network according to the embodiment of the present invention: inputting the topological structure diagram data of the space road network and the time sequence data of the exhaust concentration of the motor vehicle into a space-time diagram convolution network for completion, wherein the space-time diagram convolution network comprises: a self-attention mechanism space-time graph convolution network and an improved nearest neighbor algorithm; inputting the topological structure diagram data of the space road network into the space-time diagram convolutional network of the self-attention mechanism to obtain a space-time characteristic result of the road network, and using the space-time characteristic result in the improved nearest neighbor algorithm to complete multiple missing data of the time sequence of the exhaust concentration of the motor vehicle exhaust to obtain a final complete flow chart of the time sequence data of the exhaust concentration of the motor vehicle exhaust;

fig. 4 is a block diagram of a multi-completion method for vehicle exhaust telemetry data based on a space-time convolutional network according to an embodiment of the present invention, in which step S21: inputting the topological structure diagram data of the space road network into a self-attention mechanism space-time diagram convolution network to obtain a flow chart of space-time characteristic weight of the self-attention space-time diagram convolution of the space road network;

FIG. 5 is a block diagram of a spatio-temporal convolutional network in an embodiment of the present invention;

FIG. 6 is a block diagram of a multi-completion system for vehicle exhaust telemetry data based on a spatio-temporal convolutional network according to an embodiment of the present invention.

Detailed Description

The invention provides a multi-completion method and a multi-completion system for motor vehicle exhaust telemetering data based on a space-time convolution network, which can complete missing data of motor vehicle exhaust emission concentration telemetered by telemetering equipment in a road network. Space and time dependencies of a road network topological structure are extracted by using a space-time graph convolution and a self-attention mechanism for the first time, space-time characteristics with weights are generated, and the space-time characteristics are applied to an improved nearest neighbor method to complete exhaust emission concentration data.

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings.

Example one

As shown in fig. 1, a method for multi-complementing vehicle exhaust telemetry data based on a space-time convolutional network according to an embodiment of the present invention includes the following steps:

step S1: acquiring topological structure information of remote measurement points of a road network in a detection area and multiple missing data of the exhaust concentration of the motor vehicle remotely measured by remote measuring equipment in the road network, processing the topological structure information and the multiple missing data, and respectively constructing topological structure chart data of a spatial road network and time series data of the exhaust concentration of the motor vehicle;

step S2: inputting the topological structure diagram data of a space road network and the time sequence data of the exhaust concentration of the motor vehicle into a space-time diagram convolution network for completion, wherein the space-time diagram convolution network comprises: a self-attention mechanism space-time graph convolution network and an improved nearest neighbor algorithm; and inputting the topological structure diagram data of the space road network into the space-time diagram convolutional network of the self-attention mechanism to obtain a space-time characteristic result of the road network, and using the space-time characteristic result in an improved nearest neighbor algorithm to perform multiple missing data completion on the motor vehicle exhaust emission concentration time sequence missing data to obtain final complete motor vehicle exhaust emission concentration time sequence data.

The invention provides a multiple completion method of motor vehicle exhaust telemetering data based on a space-time convolution network, which is used for detecting the space-time dependency of a regional road network topological structure. The method not only uses the spatial dependence of the detection area road network topological structure, but also considers the data completion of the time dependence on the exhaust emission concentration. Missing motor vehicle exhaust emission data is supplemented in a novel mode of adopting space-time information of road network topology.

As shown in fig. 2, in one embodiment, the step S1: acquiring topological structure information of remote measuring points of a road network in a detection area and multiple missing data of the exhaust concentration of the motor vehicle remotely measured by remote measuring equipment in the road network, processing the topological structure data and the time series data of the exhaust concentration of the motor vehicle in the space road network, and respectively constructing the topological structure data and the time series data of the exhaust concentration of the motor vehicle in the space road network, wherein the topological structure data comprises the following steps:

step S11: acquiring topological structure information of remote measuring points of a spatial road network of a detection area, and constructing spatial topological structure chart data of the road network;

step S12: the method comprises the steps of collecting motor vehicle exhaust emission concentration multiple missing data telemetered by telemetering equipment in a road network, and constructing exhaust concentration emission time sequence input data of multiple time periods.

In one embodiment, the step S11: acquiring topological structure information of remote measuring points of a spatial road network in a detection area, and constructing spatial topological structure diagram data of the road network, wherein the method comprises the following steps:

step S111: acquiring longitude and latitude in a map and approximate distances between detection points in road network remote sensing detection nodes in a detection area, wherein the number of the remote sensing detection nodes is N, and the number of the nodes v is_iAnd v_jA geographical distance of l_i,jWherein i, j is epsilon to N;

the number of the remote sensing detection nodes is 114.

Step S112: constructing topological structure diagram data of a spatial road network, and defining a topological structure diagram G of the spatial road network as (V, E, A) according to a road network remote sensing detection node topological diagram, wherein G is an undirected graph; v is a node set for remote sensing detection of tail gas emission of the motor vehicle, | V | ═ N is the number of nodes, E is an edge set of remote sensing points, the connectivity of the nodes is represented, A ∈ R^N×NThe weight adjacency matrix of the graph G has a spatial weight adjacency matrix a with the mathematical expression of elements:

wherein link (i, j) represents node v_iAnd v_jThe road connection between them. When link (i, j) is 1, it represents the node v_iAnd v_jThere is a road connection between them; the parameter θ represents the size and sparsity of the control weight adjacency matrix.

In one embodiment, the step S12: the method comprises the steps of collecting motor vehicle exhaust emission concentration multiple missing data telemetered by telemetering equipment in a road network, and constructing exhaust concentration emission time sequence input data of multiple time periods, wherein the data comprises the following steps:

the method comprises the following steps of collecting motor vehicle exhaust emission concentration multiple missing data telemetered by telemetering equipment in a road network, and constructing exhaust concentration emission time sequence input data of multiple time periods, wherein the method specifically comprises the following steps:

step S121: hourly time series for obtaining and constructing motor vehicle exhaust concentration data

Step S122: acquiring and constructing a time-of-day sequence of motor vehicle exhaust concentration data

Step S123: acquiring and constructing weekly time series of motor vehicle exhaust concentration data

The emission concentration of the motor vehicle exhaust concentration data collected by the exhaust remote sensing monitoring system at the moment t is X_t(ii) a Motor vehicle exhaust gas concentration data according to T_pAccumulating within hours, in the embodiment of the invention, T is selected_p0.5 hour, therefore, the number of the data of the exhaust gas emission concentration of the motor vehicle per day isA plurality of; l_h、l_dAnd l_wRespectively, the lengths of the hourly, daily and weekly sequence segments are respectively truncated along the time axis and are T_pInteger multiples of; for example, assuming the target prediction interval is one frame in a Friday interval (e.g., 10:00-10:30), the hourly interval length l_h3 represents 3 frames of the three previous intervals of the time interval; length of daily interval l_d3-3 frames representing the same time interval on thursday, wednesday, tuesday; cycle time interval length of each week l_w3-3 frames representing the same time interval every friday in the last three weeks; d and w are 1 day and 1 week, respectively.

Step S124: constructing an exhaust gas concentration emission time sequence with multiple time periods according to the 3 groups of time period sequences, wherein input data is T ═ X_h,X_d,X_w}。

As shown in fig. 3, in one embodiment, the step S2: inputting the topological structure diagram data of a space road network and the time sequence data of the exhaust concentration of the motor vehicle into a space-time diagram convolution network for completion, wherein the space-time diagram convolution network comprises: a self-attention mechanism space-time graph convolution network and an improved nearest neighbor algorithm; inputting the topological structure diagram data of the space road network into an attention mechanism space-time diagram convolution network to obtain a space-time characteristic result of the road network, and using the space-time characteristic result in an improved nearest neighbor algorithm to perform multiple missing data completion on the motor vehicle exhaust emission concentration time sequence missing data to obtain final complete motor vehicle exhaust emission concentration time sequence data, wherein the method comprises the following steps:

step S21: inputting the topological structure diagram data of the space road network into a convolutional neural network for training, and adding an attention characteristic mechanism to obtain the space-time characteristic weight of the self-attention space-time diagram convolution of the space road network;

step S22: and inputting the space-time characteristic weight convolved by the attention space-time diagram and the missing data of the time sequence of the exhaust concentration of the motor vehicle into an improved nearest neighbor algorithm to obtain final complete time sequence data of the exhaust concentration of the motor vehicle.

The invention discloses a space-time graph convolution network which is composed of two parts, namely a self-attention mechanism space-time graph convolution network and an improved nearest neighbor algorithm. Inputting the topological structure diagram data of the space road network into an attention mechanism space-time diagram convolution network to obtain a space-time characteristic result of the space road network, and applying the obtained space-time characteristic result of the space road network to an improved nearest neighbor algorithm to complete multiple missing data of the motor vehicle exhaust emission concentration time sequence missing data to obtain final complete motor vehicle exhaust emission concentration time sequence data.

As shown in fig. 4, in one embodiment, the step S21: inputting the topological structure diagram data of the space road network into a convolutional neural network for training, and adding an attention characteristic mechanism to obtain the space-time characteristic weight of the self-attention space-time diagram convolution of the space road network, wherein the space-time characteristic weight comprises the following steps:

step S211: transforming the space road network topological structure diagram G (V, E, A) from the diagram to the spectrum domain to realize the graph convolution, and using a convolution kernel G_θPerforming convolution operation on the graph G to obtain a space graph convolution, and performing approximate expansion by using a Chebyshev polynomial to obtain a space graph convolution result H, wherein the mathematical expression of the space graph convolution result H is shown as a formula (2):

wherein_GWhich represents a convolution operation, the operation of the convolution,k order Chebyshev polynomial coefficients for K (K ∈ K); x represents a weight adjacency matrix a of the input graph G;λ_maxmaximum eigenvalue, I, of Laplace matrix representing matrix A_NIs the identity matrix of matrix A, and the Laplace matrix normalized form of graph GRecursive definition of chebyshev polynomials: t is_k(x)＝2xT_k-1(x)-T_k-2(x) Wherein，T₀(x)＝1，T₁(x)＝x。

Through a plurality of experiments, the embodiment of the invention adopts the K-3 Chebyshev polynomial, and can obtain better results. The order of the Chebyshev polynomial is not particularly limited, and the Chebyshev polynomial of which order is used can be determined according to actual requirements.

Step S212: using the self-attention mechanism method to calculate the weight coefficient alpha of the convolution result of the space map given by the self-attention mechanism according to the following formula (3) and obtain the attention time characteristic result H_Att＝α·H；

Wherein H_i，jAnd j is the value of the ith telemetry device telemetry value representing H, H is the number of hours of acquisition in the vehicle telemetry exhaust time series data set, d is the number of days of acquisition, and w is the number of weeks of acquisition.

Step S213: attention is paid to the time characteristic result H_AttThrough the calculation of at least one layer of time convolution neural network, the space-time characteristic result obtained by output isElements contained therein

Fig. 5 shows a schematic structural diagram of a space-time convolutional network in an embodiment of the present invention. Through multiple experiments, the 2-layer convolutional neural network adopted by the embodiment of the invention can obtain better results. The invention does not specifically limit the convolutional neural network, and can determine which structure of convolutional neural network to use according to actual requirements.

In this step, since there is an importance difference between the time-dependent features of the plurality of angles, the attention time feature is finally obtained by giving a weight coefficient to the result of the time feature using an attention mechanismCharacterization result X_Att。

In an embodiment, the improved nearest neighbor algorithm in step S22 specifically includes:

step S221: for exhaust emission data sets containing missing valuesAnd spatio-temporal feature resultsGiven improved nearest neighbor parameter k, T,And k inputting an improved nearest neighbor algorithm;

step S222: by the missing value x_i,jAs a center, wherein i belongs to N, j belongs to (h + d + w), constructing a root node which corresponds to a hyper-rectangular region containing Nx (h + d + w) space which lacks the exhaust emission data T; the space-time characteristic resultSequencing the values in the middle distance data table from small to large to obtain a sequencing Array, sequentially taking k values from the Array, and taking a space-time characteristic result h of the k values_i,jCorresponding motor vehicle exhaust emission data x_i,jAccording to the following formula (4), summing and averaging are carried out to obtain the missing value of the exhaust emission concentration of the position

Step S223: repeating the step S222 to finally obtain a multiple deficiency value exhaust emission data set with complete deficiency value complementation

Through a plurality of experiments, the embodiment of the invention adopts the nearest neighbor parameter k equal to 8, and can obtain better results. The nearest neighbor parameter value is not particularly limited, and the nearest neighbor parameter of the use value can be determined according to actual requirements.

Example two

As shown in fig. 6, the system for multi-completion of motor vehicle exhaust telemetry data based on the space-time convolution network is characterized by comprising the following modules:

a space road network data and exhaust emission data constructing module 31, configured to acquire topological structure information of road network remote measurement points in a detection area and multiple missing data of exhaust emission concentration of a motor vehicle remotely measured by remote measurement equipment in a road network, process the topological structure data and the time series data of the exhaust emission concentration of the motor vehicle of the space road network;

a space-time graph convolution network module 32, configured to input the topological structure diagram data of the space network and the time-series data of the exhaust emission concentration of the motor vehicle into the space-time graph convolution network for completion, where the space-time graph convolution network includes: a self-attention mechanism space-time graph convolution network and an improved nearest neighbor algorithm; inputting the topological structure diagram data of the space road network into an attention mechanism space-time diagram convolution network to obtain a space-time characteristic result of the road network, and using the space-time characteristic result in the improved nearest neighbor algorithm to perform multiple missing data completion on the motor vehicle exhaust emission concentration time sequence missing data to obtain final complete motor vehicle exhaust emission concentration time sequence data.

The above examples are provided only for the purpose of describing the present invention, and are not intended to limit the scope of the present invention. The scope of the invention is defined by the appended claims. Various equivalent substitutions and modifications can be made without departing from the spirit and principles of the invention, and are intended to be within the scope of the invention.